We propose to test the hypothesis that a computer programmed to interact in dialogue can provide direct assistant to patients with important, common medical problems. We have chosen urinary tract infection, sore throat, hyperlipoproteinemia and headache for our preliminary experiments because these disorders are representative of problems that require individualized attention too complex for pamphelets and books. Professional help, which is traditionally required for these problems, is sometimes unavailable and amost always expensive. We hope to demonstrate that the automated patient's assistant can be effective, thorough and inexpensive and can provide a supportive, personalized attention sometimes missing from a busy clinic. Furthermore, we plan to use the information obtained through patient-computer dialogue to learn more about the diagnosis and treatment of urinary tract infection, sore throat, hyperlipoproteinemia and headache.